Listened to Railo's Gert Franz, as part of the Railo US Tour 08, speak at the Baltimore Adobe User Group last night. And I came away with a very position impression of Railo as CFML engine. Gert spoke for almost three hours and covered a significant amount of territory. The following are what I considered some of the most interesting items from that presentation.

Video

A cfvideo tag was demonstrated that had the capability to play, convert, splice and makes images from video files in several formats including AVI, FLV and MPEG. The most interesting use demonstrated was taking image shots from several points in a video and then displaying those as thumbnails that would then jump to the corresponding spot and start playing. In conjunction was a demonstration of how to move the video processing and serving off the web server and instead use Amazon's cloud computing and storage (S3), basically pay per use utliity computing, for processing and serving respectively. This is particularly interesting because Railo will offer an server image for Amazon's cloud computing.

It should be noted however, this will not be part of the Railo open source distribution. Instead it will be a pay for extension. In part this is due to the backing library being used is a commercial product.

Virtual File Systems & Application Archives

Application server mappings and tag/function attributes/arguments can point to things other than actual operating system file systems. For instance, using a .zip file the source of files for a given mapping was demonstrated. But this can also be other things such as Amazon's S3 or an FTP server.

In addition it's possible to very easily create a sourceless application distribution through Railo's administrator. Granted there's the Adobe cfcompile, but what was demonstrated was literally defining a mapping and clicking create a Railo archive. This archive in turn can then be deployed on another Railo server by setting up a mapping to it. The implications for distributing frameworks and libraries such as Fusebox and say Coldspring are significant because the deployment now becomes only setting up a mapping to a single file rather than expanding an archive onto the filesystem.

Clustering Administration and Data Sharing

The adminstrator allow for configuring servers, things like datasources and mappings, across a cluster in one place. Also Railo has a cluster scope that when changed is propagated across other servers in the cluster, so it behaves as a shared memory space. Again this has been done before and can be accessed from Coldfusion through J2EE session sharing for instance, but this strikes me as a far more elegant solution. That said, the usual caveats about don't abuse shared memory because of all the synchronization and messaging involved applies here.

Conclusion

A disclaimer, I have not used Railo yet. But, based on this presentation I see a feature set not offered with Adobe's engine that I can realistically see uses for in the work I am doing now. If you have the opportunity to see Gert speak on this road show I highly recommend it, because he offers a very in depth discussion of the engine and its feature sets that I think any Coldfusion programmer will find very relevant, and thought provoking, to say the least. At least it was for me.